loading
loading
NIH grant 1R01NR022435-01: Discovery of Symptom Burdens and Their Associations with Health Outcomes among People Living with HIV
Institute: {'code': 'NR', 'abbreviation': 'NINR', 'name': 'National Institute of Nursing Research', 'admin_org_id': '12500052', 'admin_funding_url': ''}. PI: TRU HOANG Cao. Award: $402355.
Start: 2026-09-25T00:00:00.
ABSTRACT
In the post combined antiretroviral therapy era, symptom management is a priority in the care of people living
with HIV (PWH). For effective symptom management, determination of the symptom burden of PWH,
identification of risk factors for severe symptoms, and assessment of the impact of symptoms on relevant
outcomes are important. The ability to characterize symptom clusters in PWH allows clinicians to target multiple
symptoms that co-occur, preventing the occurrence of symptoms in the same cluster, and implementing group-
specific symptom management strategies. Symptom management concerns associations between the three
domains of symptom management, symptom experience, and health outcomes. Symptoms are potential
predictors of clinical outcomes such as mortality and quality of life. However, symptom cluster analysis is still an
emerging field in HIV research. Little is known about the interconnected nature of co-occurring symptoms and
their potential effects on clinical outcomes of PWH. Our overall long-term goal is to identify symptomatic cues of
potential adverse outcomes for timely intervention and prevention in caring for PWH. This study aims to develop
innovative methods and tools to discover concurrent symptom burden patterns and their associations with
important clinical outcomes among PWH using electronic health records (EHRs), as it would be costly and labor
and time intensive to answer these clinical questions by recruiting and surveying patients.
The first aim of this study is to develop a hybrid phenotyping method to identify HIV cases from EHRs. We
will develop a machine learning-based classifier exploiting MIMIC-IV clinical notes to identify HIV cases that do
not have matching HIV-specific medications. It will be integrated into our previously devised highly precise HIV
phenotyping algorithm that used only structured data to improve its recall. The second aim is to develop natural
language processing tools to extract HIV symptoms from clinical notes. We will adapt CLAMP and LLM-IE as
two alternative approaches to extract HIV symptoms and their associated attributes from MIMIC-IV clinical notes
among the PWH identified from the first aim. The third aim is to discover HIV symptom patterns and their
associations with clinical outcomes. We will adapt our previous symptom-based Latent Dirichlet Allocation model
to discover patterns of the HIV symptoms extracted from the second aim. The obtained symptom clusters will be
used to characterize and cluster PWH. We will compare the demographics and symptom characteristics of the
derived PWH clusters and analyze their associations with in-hospital and out-of-hospital mortality, ICU and
hospital lengths of stay, and 30-day readmission.
The project will produce innovative data driven methods and tools to advance HIV symptom research using
massive and invaluable medical data in EHRs. The new clinical findings can be translated into HIV
symptomatological insights and group-specific symptom management, intervention, and treatment strategies.
The approaches and software can be reused and adapted for symptom studies on other medical conditions.document fetched 2026-10-04 02:43